Developing with User Model and AI Library API_2 - 3.5 English

Vitis AI Library User Guide (UG1354)

Document ID
Release Date
3.5 English
To use your own models, your model framework should be within the scope of the Vitis AI Library. This section shows you how to deploy a retrained YOLOv3 Caffe model to the ZCU102 platform based on the Vitis AI Library.
  1. Download the corresponding Docker image from
  2. Load and run the Docker.
  3. Create a folder and place the float model within it on the host side, then use the Vitis AI Quantizer to quantize the model. For more details, see the Vitis AI User Guide (UG1414).
  4. Use the Vitis AI Compiler to compile the model to an xmodel file, such as yolov3_custom.xmodel. For more information, see the Vitis AI User Guide (UG1414).
  5. Create the yolov3_custom.prototxt, as shown in the following snippet.
    model {
      name: "yolov3_custom"
      kernel {
         name: "yolov3_custom"
         mean: 0.0
         mean: 0.0
         mean: 0.0
         scale: 0.00390625
         scale: 0.00390625
         scale: 0.00390625
      model_type : YOLOv3
      yolo_v3_param {
        num_classes: 20
        anchorCnt: 3
        layer_name: "59"
        layer_name: "67"
        layer_name: "75"
        conf_threshold: 0.3
        nms_threshold: 0.45
        biases: 10
        biases: 13
        biases: 16
        biases: 30
        biases: 33
        biases: 23
        biases: 30
        biases: 61
        biases: 62
        biases: 45
        biases: 59
        biases: 119
        biases: 116
        biases: 90
        biases: 156
        biases: 198
        biases: 373
        biases: 326
        test_mAP: false
    Note: The <model_name>.prototxt file is effective only when you use the Vitis AI Library API_1.

    The parameter of the model needs to be loaded and read manually by the program when using the Vitis AI Library API_2. See the Vitis-AI/examples/vai_library/samples/dpu_task/yolov3/demo_yolov3.cpp file for details.

  6. Create the demo_yolov3.cpp file. See the Vitis-AI/examples/vai_library/samples/dpu_task/yolov3/demo_yolov3.cpp file for reference.
  7. Create a file as shown below, or copy one from the Vitis AI Library demo and modify it.
    $CXX -std=c++17 -O3 -I. -o demo_yolov3 demo_yolov3.cpp -lopencv_core -lopencv_video -lopencv_videoio -lopencv_imgproc -lopencv_imgcodecs -lopencv_highgui -lglog -lxnnpp-xnnpp -lvitis_ai_library-model_config -lprotobuf -lvitis_ai_library-dpu_task
  8. Exit the Docker tool system and start the Docker runtime system.
  9. Cross-compile the program and generate an executable file called demo_yolov3.
    sh -x
  10. Create a model folder in the /usr/share/vitis_ai_library/models folder on the target side.
    mkdir yolov3_custom /usr/share/vitis_ai_library/models
    Note: /usr/share/vitis_ai_library/models is the default location for the program to read the model. You can also place the model folder in the same directory as the executable program.
  11. Copy the yolov3_custom.xmodel and the yolov3_custom.prototxt files to the target and put them in the /usr/share/vitis_ai_library/models/yolov3_custom location.
    scp yolov3_custom.xmodel  yolov3_custom.prototxt root@IP_OF_BOARD:/usr/share/vitis_ai_library/models/yolov3_custom
  12. Copy the executable program to the target board using scp.
    scp demo_yolov3 root@IP_OF_BOARD:~/
  13. Execute the program on the target board to get the following results. Before running the program, ensure that the target board has the Vitis AI Library installed, and prepare the images you want to test.
    ./demo_yolov3 yolov3_custom sample.jpg